Gabor Filter in fingerprint image enhancement

Hello, im trying to enhance fingerprint image using gabor filter. I have already estimated the orientation field and local frequency and according to this constructed the Gabor filter. I divided image into non-overlaping blocks of size WxW and applied appropriate filter into each block using convolution (conv2)...but the output image isnt good and ridges are usually broken at the boarders of blocks.

I have searched internet a lot, but didnt find answer. What am i doing wrong? And what are the correct steps in applying gabor filter on fingerprint?

Thanks :)

[David Young]

Why are you dividing the image into blocks? Normally a convolution is applied to the whole image - what is the purpose of dividing into blocks?

[Adam]

I guess because in different parts of image, fingerprint has different orientations...so I can fit into these parts a filter with same orientation.

How do I enhance image of fingerprint with filter (one orientation) if image has many different orientations?

[David Young]

In general, I think you would apply all the Gabor filters to the whole image, and then for each block select the best output to splice into the final image. An example of this sort of approach is here. There is quite a lot of related work, and the best way to get started is probably to reimplement a specific paper then make adaptions.

However, sticking with what you have done, your problem might be associated with what the convolution operator does at the block boundaries. You might get some improvement simply by doing the convolutions first, then cutting the results into blocks and putting them into the mosaic, rather than cutting the original image up first.

[Adam]

I´ll try it. thanks :)

[Sajjad Ali]

how to connect distorted lines in fingerprint through enhancement?

[keenjhar ayoob]

can you please share me the code