How can I remove a picture's watermark using Matlab's image processing toolbox?

Which image processing algorithms are simplest to be implemented using MapReduce framework?

• Though MapReduce may not be the best way to implement the algorithms used in Image Processing, just out of curiosity, which would be the simplest ones to implement if I were to try them out as a beginner.

• Answer:

  If you do not intend to parallelize the processing of one image, but simply parallelize by image, then this is a trivial application of MapReduce, and anything you can implement in Java (or call natively from Java) is possible. In this case your job would involve Mappers and no Reducers, and each Mapper would take in a single image as a record, write a single image as a record. If you mean to parallelize the processing of one image, I would suggest that it is difficult and probably not helpful to parallelize. Parallelizing image processing almost certainly means parallelizing by row, but, unfortunately just about every 2D image algorithm does not process rows completely independently. It takes some expensive acrobatics to write processes where two inputs depend on one another. And, unless the image is very, very large (>10GB) it fits in memory in a big machine that you can rent from Amazon AWS right now. An in-memory process, using multiple cores or not, will almost always beat a Hadoop-based implementation. You don't have this problem of chopping up and processing parts of the data independently, and, you can use so many more off-the-shelf libraries. So, I doubt this will be worthwhile in practice, and probably not even the best exercise to just learn about Hadoop. But if you must: the simplest algorithm you can implement on Hadoop is one that could process a huge image by row. Anything describable by a 1D kernel works, like a simple sharpening or blur filter.