Improve the clarity of the R329 development board camera

tobot · Published on 2021-10-16 22:46

Improve the clarity of the R329 development board camera [Copy link]

In previous experiments ( https://bbs.eeworld.com.cn/thread-1181419-1-1.html , https://bbs.eeworld.com.cn/thread-1182177-1-1.html ), we found that the video captured by the camera of the R329 development board was relatively blurry, and because the image information was processed, the image on the display screen took a long time, and then fed back to the display screen, resulting in an image that was not smooth. However, both image recognition and classification only require static images. Based on this idea, I thought about it in reverse. If I capture a video in advance and then extract the clearest picture from it for processing, will the effect be better? In addition to manually selecting the "clearest" picture from multiple pictures, can we also use the board to "intelligently" select the highest quality picture? Opencv usually recommends processing pictures with Laplace factors, where pictures with larger values will give people a clearer feeling.

Based on this consideration, we tried to use opencv to record several frames, and used the Laplace factor calculation method for each frame. We compared the image quality through the calculation results and recorded the image with the best quality.

After trying, it seems that reducing the camera resolution can improve the Laplace calculation results.

Here are some of my records.

Original image:

Recorded via camera:

Laplace calculation value: 41.48

Original image:

Get pictures through the camera:

Laplace calculated value: 83.62

Data processing can then be performed based on the optimal pattern.

tobot · Published on 2021-10-16 22:51

excellentpic.py (1.29 KB, downloads: 0, 售价: 1 分芯积分)

This is a script I wrote in python. It loops digitally, takes 500 pictures, and records the best picture.

tobot · Published on 2021-10-16 23:14

Cut the original picture to 4:3 and reduce the size to 320*240, and write code to calculate the Laplace value.

The Laplace value in the above image is 1365.78, which is much larger than the number seen from the camera, indicating that it is much clearer than that image.