Download to read offline
![RasterExpressionEvaluator
● Simple value assignment
○ Red8 = 255
● Accepts values from attributes
○ Green8 = A:_multiplier
● Evaluates expressions on pixel values
○ Blue8 = A[2]*1.1
● Use conditions, which can be simple or complex and nested
○ Blue8 = if(A[2]>127,255,0)
○ UInt8 = if(A[0]==0&&A[1]==0&&A[2]==0,0,if(A[0]<127&&A[1]<127&&A[2]<127,1,2))](https://image.slidesharecdn.com/makingsenseofrasterfrombitdepthtobetterworkflows-dec102025-251210180941-405f9ed5/75/Making-Sense-of-Raster-From-Bit-Depth-to-Better-Workflows-45-2048.jpg)
![Understanding Raster Expressions
● If Raster A, band zero is less than 255, then set the current cell to 200, else leave it as 255.
● If Raster A, band 0, 1 & 2 are greater than or equal to 230, then set the current cell to 0, else
leave it as 1
if(A[0]<255,200,255)
if(A[0]>=230&&A[1]>=230&&A[2]>=230,0,1) We will use
this in a demo
later today!](https://image.slidesharecdn.com/makingsenseofrasterfrombitdepthtobetterworkflows-dec102025-251210180941-405f9ed5/75/Making-Sense-of-Raster-From-Bit-Depth-to-Better-Workflows-46-2048.jpg)
Uploaded bySafe Software
Making Sense of Raster: From Bit Depth to Better Workflows
Raster data plays a powerful role in mapping, analysis, and visualization – but it can also be intricate, with layers of detail that shape how your workflows perform. In this webinar, we’ll unpack core raster concepts by exploring what they are, how they work, and why they matter: helping you troubleshoot confidently and design more effective workflows. From supplier data mismatches to compressing citywide imagery or preparing files for web maps, we’ll connect theory to practice and demonstrate how understanding these principles leads to better results. During the hour, we’ll explore: -Core raster concepts and terminology -Managing raster interoperability -Techniques for organizing and optimizing raster data Walk away with a clearer conceptual grasp of raster data and the confidence to make informed decisions in FME -- so you can stop guessing and start transforming with understanding.
More Related Content
Similar to Making Sense of Raster: From Bit Depth to Better Workflows
![Vibe Coding vs. Spec-Driven Development [Free Meetup]](https://cdn.slidesharecdn.com/ss_thumbnails/vibecodingvsspecdrivendevelopment-251209105622-43f455e7-thumbnail.jpg?width=640&height=640&fit=bounds)
Making Sense of Raster: From Bit Depth to Better Workflows
- 1. Making Sense ofRaster: From Bit Depth to Better Workflows
- 2. Kailin Opaleychuk Technical Support Specialist, FMEForm Crystal Wang Technical Support Specialist, FME Form Pierre Koelich Technical Support Specialist, FME Form Insert Headshot Here. Insert Headshot Here. Insert Headshot Here. Meet the Presenters
- 3. Welcome to Livestorm. Afew ways to engage with us during the webinar: Audio issues? Click this for 4 simple troubleshooting steps.
- 4. Agenda 1 Introduction 2 KeyTerminology 3 Raster Bands & Palettes 4 Compression & NoData 5 Rasters: Formats, Properties, Transformers 6 FME and AI 7 FME for Raster Processing 8 Potential Workflow Scenario 9 Conclusion 10 Resources & Next Steps 11 Q&A
- 5.
- 6. Understanding the structureof rasters unlocks the ability to manage and transform them effectively in FME
- 7. Why Understanding RastersMatters Rasters Are Everywhere ● Imagery, terrain, sensors, regular photos, paper scans, etc. ● Power workflows from web mapping to analysis Understanding Leads to Better Results ● Diagnose issues correctly and choose the right tools for our problems ● Create raster workflows in FME with confidence ○ Be able to adapt creatively when data doesn’t behave as expected
- 8.
- 9. 1858 First Aerial Photo Early 20th century AnEarly Precursor to Multispectral Imagery WWI & WWII Airborne Photography Became Systematic 1960s The first Earth-observing Satellites 1980s and 1990s Digital sensors 2000s High-res satellites and near-real-time data streams. Snapshot History of Remote Sensing
- 10. 1950s Digital imagery & raster displays emerge 1960s FirstRaster GIS concepts 1970s Analytical raster GIS takes shape 1980s Map Algebra formalized 1990s–2000s Remote sensing meets desktop GIS 2010s–Now Cloud & AI raster era Snapshot History of Rasters Credit: R. Kirsch/NIST Credit: C.Steinitz Credit: Esri ARC News Credit: GISGeography
- 11. Poll: What’s your current experiencelevel with raster data?
- 12.
- 13. ● Workflow transformedGeoTIFFs into PNGs, written in the XYZ web map tiling scheme ○ Tiling scheme selected in WebMapTiler depends on your data destination ● Note: the .wld files were unnecessary for XYZ background tile maps Demo Summary
- 14. FME simplifies raster complexity,turning technical barriers into streamlined solutions.
- 15. The only All-Data,Any-AI Platform. FME Form FME Flow Data Movement and transformations (“ETL”) workflows are built here. Brings life to FME Form workflows FME Flow Hosted Safe Software managed FME Flow fme.safe.com/platform FME Enterprise Integration Platform Safe & FME FME Realize Experience data in real world context, in real time.
- 16. With 500+ supporteddata integrations in FME. Unrivalled Data Support GIS CAD Database XML Raster 3D BIM Web Point Cloud Cloud Big Data IOT Graph BI Indoor Mapping AR/VR Generative AI Cloud Native Tabular
- 17.
- 18. Raster, Image, Imagery ●Raster — a data structure, grid (DEM, orthophoto, NDVI, scanned topo maps, bit masks - anything in a form of columns and rows of pixels) ● Image — raster representation of what we can see directly or with instruments, even outside our visible range - RGB photos, multispectral, radar, heat, MRI - anything reflected/emitted ● Imagery — collection of images, usually in specialized industries (Satellite imagery, medical imagery) Credit: Silvia Mantovanini & the GLEAM-X Team
- 19. What Are RasterCells, What Are Pixels ● A raster cell — a unit of a raster grid representing measured or computed value ● A pixel — a cell displayed on screen, visual representation of the cell ● Cell Properties - size (x and y are usually equal but they don’t have to be), origin, extents, value(s), nodata value
- 20. What Are RasterBands ● A raster band is a two-dimensional array of cells representing a single* measured or derived value. ● Properties of Raster Bands — name, interpretation (data type and bit depth), Nodata value, Number of Palettes ● Possible Types — UINT, INT, Real, Red, Green, Blue, Alpha, Gray ● Possible Bit Depth — 8, 16, 32, 64 (legacy rasters can be 1-bit, JPEG bands can be 12-bit) ● Nodata — any number, but usually some extreme in the range (0, 255,-32768, -9999) r.
- 21. How bands work ●We often hear bands referred to as layers, but this can be misleading ● Imagine you have three paint buckets, each containing a different color. ● Instead of painting separate layers, these buckets pour different amounts of their color into the same box, where that box represents a single pixel.
- 22.
- 23. Why Understanding Bands isImportant ● Format compatibility ● Data ranges and interpretation influence size ● Proper data assembly ○ Right order can control how data looks ○ FME does not care about order ● Analysis (NDVI, NDWI, SAVI, etc) often requires specific bands ● Putting bands together requires identical dimensions and other properties
- 24. Scenarios Where BandKnowledge Really Matters ● 4-band PNG → JPEG? Nope (JPEG = 3 bands only) ● Landsat 8 natural color: 4-3-2 (in this order!) ● NDVI: (NIR – Red) / (NIR + Red) ● Mosaicking: 3-band ≠ 4-band ● Bit depth: 16-bit PNG ≠ “16-bit JPEG” (JPEG tops at ~12-bit)
- 25. What Are RasterPalettes A palette (or color table, or LUT) is a lookup list that assigns values to each cell value. ● Used with single-band rasters to display colors (or strings!) without storing values directly. ● Each cell stores only an index, and the palette defines what that index looks like. ● A band can have multiple palettes ● Typical in classified data, scanned maps, or legacy 8-bit imagery. Palette Key Interpretations Palette Values UInt8 UInt16 UInt32 (minimum 16 keys up to the interpretation type) RGB24 / RGBA32 — standard 8-bit color tables RGB48 / RGBA64 — 16-bit, smoother gradients / analytical use GRAY8 / GRAY16 — grayscale tables (intensity, elevation) STRING — maps numeric values to text labels (vs colors) (🟦 0 = Water 🟩 1 = Forest 🟫 2 = Urban ⬜ 3 = Bare soil 🟪 4 = Wetlands)
- 26. Palette Example RGBA32 0 174,225,124,255 1172,155,103,255 2 82,15,149,255 3 67,171,171,255 STRING 6 0 Water 1 Forest 2 Field 3 Marsh
- 27. Why Understanding Palettesis Important ● Storage efficiency: smaller file sizes ● Human Readable: easy to edit ● Data Classification: each class is defined by color or label ● Format support: required for GIFs ● Easy to apply consistent look across datasets ● Paletted rasters can break workflows
- 28.
- 29. What is Compression ●Compression - reducing file size by storing data more efficiently ● Can be lossless or lossy ● Works on cell patterns (repeated values, gradients, blocks of uniform color) ● Part of storage, not data
- 30. Why Compression Matters ●File size & performance — smaller, faster to read and transfer. ● Storage cost — critical for cloud workflows and FME Flow servers ● Compatibility — some formats require or forbid certain methods (e.g., JPEG only for 8-bit). ● Quality vs accuracy — lossy compression can blur or distort analytical rasters. ● Workflow efficiency — tiled + compressed rasters (like COG) open instantly and stream smoothly. ● Recompression can slow processing or degrade quality. AAAABBBCCABAAABB 4A3B2CAB3A2B
- 31. Choosing the RightCompression Method 1. What type of raster do you have? ● Imagery (photos) → JPEG, JPEG2000, WebP, ECW, MrSID ● Analytical data (DEMs, classified rasters) → LZW / DEFLATE ● Maps with transparency → PNG ● 1-bit masks, FAX Documents → CCITT Group 4
- 32. Choosing the RightCompression Method 2. Do you need exact pixel values? ➡ Yes → Use Lossless (LZW, DEFLATE, PNG) ● Great for analysis, DEMs, classifications. ➡ No → Use Lossy (JPEG, JP2, WebP, ECW) ● Best for imagery, backgrounds, visualization.
- 33. Choosing the RightCompression Method 3. Are you working with very large rasters or cloud workflows? ➡ Yes → Use COG ● Same formats, but optimized for streaming. ● Great for huge mosaics, web maps, cloud storage.
- 34. Choosing the RightCompression Method Format Lossless / Lossy Best For Why Choose It GeoTIFF (LZW/DEFLATE) Lossless DEMs, classifications Safe for analysis; preserves exact values GeoTIFF (JPEG) Lossy Orthoimagery Small file sizes; visually good PNG Lossless Maps, transparency Works well for cartography; limited precision COG Both Cloud + large rasters Fast streaming; ideal for web/cloud JPEG2000 / WebP/ ECW / MrSID Lossy / near-lossless Imagery Good quality-to-size ratio CCITT Group 4 Lossless 1-bit masks Super small for black/white scans
- 35.
- 36. What Is NoData? NoData= “not a real value.” ● A special value marking pixels with no valid data (outside coverage, missing measurements, etc.). ● The pixel still exists in the grid, but is ignored in display and analysis. ● Stored as metadata, though some formats represent it explicitly. ● Each band can have its own NoData value, depending on the format/workflow. ● Typical values: 0, 255, -9999
- 37. Why NoData Matters ●Raster calculations — nodata destroys data. Remove or handle it separately (using @isnodata()) ● In Mosaicking, defines what we’ll see in the results ● Reprojecting — nodata allows avoiding black cells over the images ● Visualization — controls what is transparency (similar, but not identical to Alpha band) ● Vectorization — allows extracting only meaningful data ● Storage — allows better compression Nodata is for logic and analysis Alpha is for visualization
- 38.
- 39. Poll: What best describesyour current raster workflows?
- 40.
- 41. Raster Formats The rasterformat is a container that defines how the following components are stored: ● Cell value encoding ● Metadata, georeferencing, compression handling ● Band, color tables, tiling organization The format defines what combination of those components can exist — and FME’s strength is that it can translate between them even when formats don’t align.
- 42. Key Raster Properties ●Raster Interpretation (composed from band interpretations) ○ Number of bands ○ Data types ● Palettes ● Nodata ● Resolution (dimensions — number of rows and columns). ○ Some format can contain multiple resolutions, raster pyramids or overviews) ○ Formats may have dimension limits ● Compression options ● Georeferencing support ● Metadata ● Interleaving (rarely matters these days) ● Cloud readiness
- 43. Raster Transformers Bands RasterSelector,RasterBandInterpretationCoercer, RasterBandNodataSetter, etc. Palettes RasterBandSeparator, RasterPaletteRemover, RasterPaletteGenerator, RasterPaletteExtractor, etc. Extracting Properties RasterStatisticsCalculator, RasterBandPropertyExtractor, RasterPropertyExtractor, etc. Pixel Operations/Analysis RasterExpressionEvaluator, RasterCellValueCalculator, RasterConvolver, etc. Rasters RasterMosaicker, Tiler, WebMapTiler, RasterPyramider, RasterAspectCalculator, etc. Vectors TINGenerator, Clipper, RasterDEMGenerator, SurfaceDraper, RasterExtentsCoercer, etc. 3D AppearanceSetter, AppearanceRemover, AppearanceStyler, SharedItem…, etc. Point Clouds PointCloudCombiner, PointCloudOnRasterComponentSetter Attributes RasterReplacer, RasterGCPSetter, AttributeFileReader/Writer, ChartGenerator, etc. Web/AI/ML GoogleVisionConnector, AzureAIVisionConnector, OpenAIVisionConnector, HTTPCaller
- 44. RasterExpressionEvaluator Simple pixel calculator ●Changes rasters by applying one or many expressions per pixel per band ● Can take >1 rasters with identical dimensions as inputs ● Outputs a new raster that may look similar to the original or totally different in the number of bands and their interpretations ● Can perform operations such as: ○ Masking, Nodata preparation for vectorization, pansharpening, normalization, brightness correction, mixing band values, etc
- 45. RasterExpressionEvaluator ● Simple valueassignment ○ Red8 = 255 ● Accepts values from attributes ○ Green8 = A:_multiplier ● Evaluates expressions on pixel values ○ Blue8 = A[2]*1.1 ● Use conditions, which can be simple or complex and nested ○ Blue8 = if(A[2]>127,255,0) ○ UInt8 = if(A[0]==0&&A[1]==0&&A[2]==0,0,if(A[0]<127&&A[1]<127&&A[2]<127,1,2))
- 46. Understanding Raster Expressions ●If Raster A, band zero is less than 255, then set the current cell to 200, else leave it as 255. ● If Raster A, band 0, 1 & 2 are greater than or equal to 230, then set the current cell to 0, else leave it as 1 if(A[0]<255,200,255) if(A[0]>=230&&A[1]>=230&&A[2]>=230,0,1) We will use this in a demo later today!
- 47.
- 48. FME and AI Howcan AI assist with raster workflows? ● AI-assisted development. ○ Recommends libraries, writes code, helps debugging ● AI-guided expression writing ○ Creates expressions for REE or matrices and weights for RasterConvolver ● AI-driven image creation ○ Texture generation ○ Retouching ○ Object detection ○ OCR
- 49. Raster Encoding ● Unstructureddata can be encoded into BLOBs (Binary Large Objects) ● BLOBs are chunks of binary data that can hold: ○ Images, audio, video, and more ● BLOBS are used by LLMs, APIs, XML, and Databases to store attachments
- 50.
- 51. 7 FME for Raster Processing Choosingthe Best Approach: When FME Is the Right Fit
- 52. Where FME fits ●If it is about what the data does and where it goes, use FME ○ Tiling, clipping ○ Mosaicking ○ Reprojection ○ Band Manipulation ○ Format Conversion ○ Rasterization ○ Vectorization ○ Per-pixel processing ○ Overlaying ○ Texturing geospatial data
- 53. Limitations of FME ●Image Editing and Visualization ○ Manual photo-style editing (brightness, contrast, curves, retouching, masking).Selective or brush-based adjustments. ○ Real-time interactive visualization ○ Layer-based editing ○ Artistic effects ○ Batch image styling ● Specialized Analytical Workflows ○ Advanced remote-sensing analytics (e.g., atmospheric correction, supervised classification, spectral unmixing). ○ Machine learning–based image analysis (use third-party tools via API or CL instead) ○ Terrain modeling beyond standard hillshade or slope (e.g., network modelling, flow accumulation). Though FME itself doesn’t perform these advanced analytical functions, it does allow you integrate these tools into your workflows!
- 54.
- 55. Scenario ● Orthophoto datareceived; needs conversion to web map tiles ● Rasters contain irregular white areas with non-uniform pixel values ● To prevent artifacts in the final web map, standardize white areas and make them transparent, before generating the tiles.
- 56.
- 57. ● We usedthe raster principles learned in this webinar to clean up a raster dataset, integrate with vector data, and optimize space in the creation of web map tiles. ● FME has a wide variety of tools that can be used to solve raster-related problems, all while allowing you to integrate other formats into the same workflow. Demo Summary There are many ways to solve the same problem in FME! We could also have used the RasterCellValueReplacer instead of the RasterExpressionEvaluator to get rid of the white areas in the image.
- 58.
- 59. Summary ● After ahistory of raster data, we learned the basics of rasters as a data format. ● We defined and reviewed key raster terminology ● Lastly, showcased how FME can create and improve raster processing workflows. X-ray: NASA/CXC/SAO; IR & UV: NASA/JPL-Caltech; Optical: NASA/STScI
- 60. 30+ 30K+ 128 140+ 25K+ years of solvingdata challenges FME Community members countries with FME customers organizations worldwide global partners with FME services 200K+ users worldwide 200K+ users worldwide
- 61. All Data. AnyAI. All Data Velocities Batch (ETL, Reverse ETL, ...) Event ( BPA, RPA, ...) Stream All Data Locations Any Cloud On-premises Hybrid Edge Containers Embedded Mixed All Data Types Unstructured Structured Spatial APIs Web Apps … Any AI Technology OpenAI Amazon Bedrock Google Gemini Ollama Deepseek Composite
- 62.
- 63. More Raster Resources ●Getting Started with cloud-native geospatial formats ● Working with Raster and Imagery Data in FME ● Image Compression in FME ● Vector & Raster Data: Converting Geometry models Webinar: Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
- 64. Get our Ebook SpatialData for the Enterprise fme.ly/gzc Guided learning experiences at your fingertips academy.safe.com FME Academy Resources Check out how-to’s & demos in the knowledge base support.safe.com Knowledge Base Webinars Upcoming & on-demand webinars safe.com/webinars
- 65. ClaimYour Community Badge& Dive into the new Community! ● Get community badges for watching webinars ● community.safe.com ● Today’s code: J20V81 Join the Community today! Next Steps
- 66.
- 67. ThankYou Recap of NextSteps 1 Follow us on LinkedIn! 2 Contact us 3 Experience the FME Accelerator Please fill out our webinar survey