Support hotline：13923733903

【Abstracts】 The invention discloses a method for judging the pollution of a camera module image, comprising the following steps: (1) dividing the entire image data into A and B regions; (2) dividing the A and B regions according to the Bayer pattern; Four blocks of Gr, R, B, and Gb are averaged according to the respective positions to generate a new block of data blocks; (3) The block of data blocks is the original data, and the point of sudden change uses the average of the surrounding data. Substitution; (4) Substituting N*N data points as partitioning elements, performing median value calculation and replacing the splitting element with a median value in place of the original data; (5) Judging dead pixels and effective points to form a new data result chart (6) Find consecutive bad spots and separate bad spots, and record relevant data information; (7) Determine whether there is dirt; (8) Repeat the search to determine the dirt by changing the size of the segmentation element. The present invention can effectively reduce the variation caused by the color temperature, and can detect the image defect of more than 2% variation.
[patent description] camera module image dirt determination method

【Technology】
[0001] The present invention relates to a camera module image quality detection method.

【Background technique】
[0002] At present, in the field of CMOS camera module testing, determination of defective images, spots, water lines, bubbles, etc. has always been a difficult problem. Each module production manufacturer has its own determination method. In general, the luminance ratio is More than 5% of the defects can basically be tested. There are few misjudgments and missed judgments on the production line. However, the brightness ratio is less than 29Γ5%, such as the outline of the air bubble, slight scratches, and slight watermarks. This is often impossible to detect, even if it is barely detected. Due to a lot of misjudgments, it is impossible to import production lines into production.