Make writing a habit together! This is my first day to participate in the “Gold Digging Day New Plan · April More text challenge”, click to see the details of the activity.

I. Introduction to fog removal in dark channel

Based on dark channel prior knowledge, this chapter proposes an image defogging algorithm, which separates the fogged image into structure layer and texture layer, to solve the problems of image texture details loss and unclear edge contour after the traditional dark channel prior defogging algorithm.

When defogging the structural layer, a dark channel based on superpixel is proposed, which divides the image into different non-local regions and the transmittance within each region is consistent. Aiming at noise amplification and image edge depth jump caused by traditional transmittance optimization method based on local gradient structure, a transmittance optimization method based on non-local regularization model was proposed, and the transmittance of dark channel prior knowledge failure region was optimized by tolerance mechanism. In the process of texture layer optimization, a mask indicating texture region is established to retain the high frequency information in the mask and filter out the small noise outside the mask. Finally, the optimized texture layer and structure layer are recombined to obtain the final fog-free image. The algorithm flow chart is shown in Figure 3 below.

Ii. Structural layer defogging stage

2.1 Structural layer 2.2 Obtaining the initial projective rate of superpixel dark Channel 2.3.1 Optimizing the initial transmittance by non-local region regularization model 2.3.2 Introducing tolerance mechanism to optimize the transmittance of bright region 2.4. Structural layer after mist removal

3. Texture layer optimization stage

3.1 Texture Layer 3.2 Build a mask indicating the texture area 3.3 Optimize the texture in the mask 3.4 Optimized texture layer

Iv. Overall steps

First step: input image second step: image decomposition third step (1) : structure layer defogging stage third step (2) : texture layer optimization stage fourth step: image reorganization fifth step: output image