The principle and python construction of Gaussian pyramid and Laplace pyramid - Programmer Sought
"Image Blending Using Laplacian Pyramid" is a technique commonly used to seamlessly blend two images together. It involves the use of a Laplacian pyramid, which is a multi-scale representation of an image, allowing smooth blending at different levels of detail. Here is a detailed description of the process:
Building a Laplacian Pyramid: A Laplacian pyramid is built by creating a sequence of images with varying levels of detail. First create a Gaussian pyramid for each input image. Gaussian pyramids are built by repeatedly applying a smoothing filter, such as a Gaussian filter, and downsampling the image. Each layer of the Gaussian pyramid represents an image of a different scale.
Generating Laplacian Pyramids: After building a Gaussian pyramid, you can generate a Laplacian pyramid by taking the difference between each level of the Gaussian pyramid and its upsampled version. An upsampled version can be obtained by applying a smoothing filter and then upsampling the image.
Blending Pyramids: Now that you have created Laplacian pyramids for the two input images, you can blend them together. The lowest level of the pyramid (coarsest scale) is first blended by simply averaging the corresponding pixel values. Then, for each level of the pyramid, the corresponding images are blended by weighting the pixels from each pyramid according to the blend mask.
Reconstructing the blended image: After blending the pyramids, you can reconstruct the final blended image by upsampling and combining the blended pyramid levels back together. Starting from the best ratio of the blended pyramid, iteratively upsample and add the next layer of the blended pyramid until the original image size is reached. (
- Finally, the fused image is reconstructed from the new Laplacian pyramid by upscaling and summing each layer.
）