panorama

use ezSIFT to glue photos together

port ezSIFT to javascript with emscripten

build_ezsift_with_em.sh is used to generate ezsift.js

ezsift.cpp is main source file for ezsift.js to export APIs from ezSIFT (much code copied from feature_extract.cpp and image_match.cpp)

Currently we support APIs:

• Module._createImage(w, h) -> img to create a 8bit image
• Module._disposeImage(img) to release memory alloc
• Module._putPixel(img, x, y, val), where 0 <= val < 256
• Module._getPixel(img, x, y) -> val
• Module._extractFeature(img) to extract feature into an internal list rKeypoints
• Module._backupKeypoints(index), where 0 <= index < 10, to save current internal rKeypoints into a cell at index.
• Module._restoreKeypoints(index), to load the saved list assigning to internal rKeypoints
• Module._getKeypointsN() to get current the number of key points in rKeypoints
• Module._keypointBegin() to assign the first key point to cursor in rKeypoints
• Module._keypointNext() to get next key point in rKeypoints
• Module._keypointHasNext() to check if has next key point
• Module._keypointOctave() _keypointLayer() _keypointR() _keypointC() _keypointScale() _keypointOri() _keypointDescriptor(index), where 0 <= index < 128 to get different key point attributes
• Module._keypointsMatch(indexA, indexB), where 0 <= indexA, indexB < 10, to match to saved key point list in 2 cells; save to internal list rMatch
• Module._getMatchN() to get the number of matched key points in rMatch
• Module._matchBegin() to assign the first matched key point to cursor in rMatch
• Module._matchNext() to get next matched key point in rMatch
• Module._matchHasNext() to check if has next matched key point
• Module._matchR1() _matchC1() _matchR2() _matchC2() to get matched key point position (c1, r1) -> (c2, r2); in ezSIFT, c represents x meanwhile r represents y in real image

after get ezsift.js,

• put into dist folder;
• create a folder local and put 1.png 2.png into the folder;
• run python -m SimpleHTTPServer or python3 -m http.server
• open http://127.0.0.1:8000/test.html (canvasContext.getImageData() do not work when directly open test.html in file protocol due to cross-origin)

wait for seconds, you will see a canvas drawing matched points between 2 images.