Image Comparison Techniques with Java -
i'm looking several methods compare 2 images see how similar are. plan have percentages 'similarity index' end-result. program outline this:
- user selects 2 images compare.
- with button, images compared using several different methods.
- at end, each method have percentage next indicating how similar images based on method.
i've done lot of reading lately , of stuff i've read seems incredibly complex , advanced , not me year's worth of java experience. far i've read about:
the fourier transform - im finding rather confusing implement in java, apparently java advanced imaging api has class it. though i'm not sure how convert output actual result
sift algorithm - seems incredibly complex
histograms - easiest out of mentioned far
pixel grabbing - seems viable if theres considerable amount of variation between 2 images doesn't it's going produce sort of accurate result. might wrong?
i have idea of pre-processing image using sobel filter first, comparing it. problem actual comparing part.
so yeah i'm looking see if has ideas comparing images in java. hoping there people here have done similar projects before. want input on viable comparison techniques arent hard implement in java.
thanks in advance
- fourier transform - can used efficiently can compute cross-correlation, tell how align 2 images , how similar are, when optimally aligned.
- sift descriptors - these can used compare local features. used correspondence analysis , object recognition. (see surf)
- histograms - normalized cross-correlation yields results comparing images on global level. since comparing color distributions end declaring outdoor scene lots of snow similar indoor scene lots of white wallpaper...
- pixel grabbing - no idea is...
you can overview this paper. field might content based image retrieval (cbir).
sorry not being java specific. hth.
Comments
Post a Comment