Abstract: Because traditional Gaussian filter often blurs the edge structure in images,an improved Gaussian filtering algorithm based on non-regular area was proposed in the paper.The algorithm based on Gaussian filtering breaks traditional thought of adopting fixed window,adaptively constructs non-regular Gaussian mask region through analyzing the self-similarity of texture.The rationality of weighted values distributed by Gaussian function is further improved to maintain rich texture as the potential noise or lower similar pixels are rejected in non-regular areas.Experimental results show that the proposed algorithm makes a better balance between enhancing image details and denoising,and is better than traditional Gaussian filtering and other methods.

Key words: Gausian filtering,Non-integer region,Image denoising