Local sharpness failure detection of camera module lens based on image blur assessment

Abstract

Videos and images have been widely used, and the requirements for camera imaging quality are getting higher and higher. At present, most methods of camera lens sharpness testing are divided into five areas: upper left, lower left, upper right, lower right, and middle. The test results of each area are used to approximate the overall sharpness level of the camera lens. The local sharpness failure of the camera lens cannot be solved by these methods. Because of this limitation, we proposed the idea of indirectly reflecting the local sharpness of the lens according to the image blur detection, and develop an image blur assessment method based on intensity and derivative (IDD). It can visualize the degradation process of camera sharpness from center to edge and the location of local failure. We demonstrate the feasibility and accuracy of the method through a case, as well as comparison to other methods.

Appendix 1

In Section 4, we compare the test results of seven reference-free image blur detection methods. The calculation process of the seven methods is as follows:

• 1. Sobel:

Image Src convolute with 3*3 sobel operator. The average value after convolution is taken as the index of blur detection. The smaller the value, the more blurry the image.

$$ {\displaystyle \begin{array}{c}G=\frac{1}{2}\times \left({G}_H+{G}_V\right)\\ {}=\frac{1}{2}\times \left(\left[\begin{array}{ccc}-1& 0& 1\\ {}-2& 0& 2\\ {}-1& 0& 1\end{array}\right]\ast Src+\left[\begin{array}{ccc}-1& -2& -1\\ {}0& 0& 0\\ {}1& 2& 1\end{array}\right]\ast Src\right)\end{array}} $$

(19)

Where G_H is the horizontal blur degree, G_V is the vertical blur degree.

• 2. Tenegrad:

Similar to Sobel method. The smaller the value, the more blurry the image.

$$ {\displaystyle \begin{array}{c}G=\sqrt{G_H^2+{G}_V^2}\\ {}=\sqrt{{\left(\left[\begin{array}{ccc}-1& 0& 1\\ {}-2& 0& 2\\ {}-1& 0& 1\end{array}\right]\ast Src\right)}^2+\Big(\left[\begin{array}{ccc}-1& -2& -1\\ {}0& 0& 0\\ {}1& 2& 1\end{array}\right]\ast Src}\Big){}^2\end{array}} $$

(20)

• 3. Laplacian:

Using Laplacian to detect image blur is to conclute the imageSrcwith Laplacian operator:

$$ {\displaystyle \begin{array}{c}L=\frac{1}{2}\times \left({L}_H+{L}_V\right)\\ {}=\frac{1}{2}\times \left(\left[\begin{array}{ccc}-1& 2& -1\\ {}-1& 2& -1\\ {}-1& 2& -1\end{array}\right]\ast Src+\left[\begin{array}{ccc}-1& -1& -1\\ {}2& 2& 2\\ {}-1& -1& -1\end{array}\right]\ast Src\right)\end{array}} $$

(21)

• 4. Brenner:

• 5.
  $$ {B}_H=\sum \limits_x\sum \limits_y\mid f\left(x,y+2\right)-f\left(x,y\right)\mid $$

  (22)
• 6.
  $$ {B}_V=\sum \limits_x\sum \limits_y\mid f\left(x+2,y\right)-f\left(x,y\right)\mid $$

  (23)

Where B_H is the horizontal blur degree, B_V is the vertical blur degree. The smaller the value, the more blurry the image.

• 7. Gray-scale variance:

This method only has the overall blur indexD:

$$ GV=\sum \limits_x\sum \limits_y\left(|f\left(x,y\right)-f\left(x,y-1\right)|+|f\left(x,y\right)-f\left(x+1,y\right)|\right) $$

(24)

• 8. Energy gradient:

Similar to method 5. This method also only has the overall blur indexE:

$$ E=\sum \limits_y\sum \limits_x\left({\left|f\left(x+1,y\right)-f\Big(x,y\Big)\right|}^2+{\left|f\Big(x,y+1\left)-f\right(x,y\Big)\right|}^2\right) $$

(25)

• 9. Vollath:

• 10.
  $$ V=\sum \sum f\left(x,y\right)\ast f\left(x+1,y\right)-M\ast N\ast {\mu}^2 $$

  (26)

Where M, N is the size of image, and μ is the average of image pixels.