Normalization is a step in image processing that is used to reduce lighting and contrast effects, significantly increasing the accuracy of the entire solution. The face detection algorithm proposed by Viola-Jones popularized the image normalization process. The basis of this process consists in obtaining the integral image and its square integral. This process requires a greater amount of computational resources, being then avoided, particularly in embedded systems, even if it compromises the result of the solution. In this article we propose a way to implement the image normalization process from the integral of that image already stored in memory, without increasing the amount of external memory, accessing each value of integral stored only twice.

归一化是图像处理中的一个步骤，用于减少光照和对比度效果，显着提高整个解决方案的准确性。Viola-Jones 提出的人脸检测算法普及了图像归一化过程。该过程的基础在于获得积分图像及其平方积分。这个过程需要更多的计算资源，然后被避免，特别是在嵌入式系统中，即使它损害了解决方案的结果。在本文中，我们提出了一种从已经存储在内存中的图像的积分来实现图像归一化过程的方法，而无需增加外部存储器的数量，只访问存储的每个积分值两次。