1024*1024 = 1,048,576 1 billion pictures is 1000,000,000 Each pixel corresponds to about 1000 pictures. You can use the sampling method. Each pixel randomly selects N pictures from 1000 pictures. , then randomly select M pixels in each picture, obtain the average value as a pixel of the thumbnail, and obtain 1024*1024 pixels of the thumbnail in turn. Can be implemented using opencv. However, if these 1 billion pictures are random, the result will be close to a picture with a grayscale of 0.5. Because according to the central limit theorem, the variance of each pixel of the thumbnail tends to 0. You have to ensure that the combination of 1 billion pixels has structural characteristics.
Data compression will calculate the pixel size of each image on the final image, then compress 1 billion images to the corresponding size, and finally call ImageMagick for splicing
I’m not sure about the principle of the thumbnail generation algorithm, or can you see if you can parallelize the algorithm and do it in a distributed way?
![[Web front-end] Node.js quick start](https://img.php.cn/upload/course/000/000/067/662b5d34ba7c0227.png)
1024*1024 = 1,048,576
1 billion pictures is 1000,000,000
Each pixel corresponds to about 1000 pictures. You can use the sampling method. Each pixel randomly selects N pictures from 1000 pictures. , then randomly select M pixels in each picture, obtain the average value as a pixel of the thumbnail, and obtain 1024*1024 pixels of the thumbnail in turn.
Can be implemented using opencv.
However, if these 1 billion pictures are random, the result will be close to a picture with a grayscale of 0.5. Because according to the central limit theorem, the variance of each pixel of the thumbnail tends to 0. You have to ensure that the combination of 1 billion pixels has structural characteristics.
Open PS and draw a picture directly with maximum efficiency
Again, Alibaba Cloud OSS comes with image processing functions. You can take a look at the API.
Data compression will calculate the pixel size of each image on the final image, then compress 1 billion images to the corresponding size, and finally call ImageMagick for splicing
I’m not sure about the principle of the thumbnail generation algorithm, or can you see if you can parallelize the algorithm and do it in a distributed way?
Why is there such a demand? 100,000 such small pictures cannot take up even one pixel in each picture, right?
It feels like the memory will fail
1 billion, what is the concept? It’s big data, right?
1 billion pictures! ! I want it fast! ! Looking for solutions