I have a video file.I have converted this video to set of Image frames with Xuggler now I have to read each frame of image and needs to do some processing. Say we got 2000 such images (i am using Xuggler- This converts Video to Image and then Image to Video). Now say in first Image i have to draw a rectangle on the given co-ordinates. Once all processing of image gets over then again all image gets converted to video by Xuggler tool.
As video file is very huge like 5TB and more. I want to do all Image processing work on Hadoop mapreduce. I found one API HIPI which is dedicated to Image processing in the Hadoop.
However HIPI does not support to add MetaData information. I need some metadata information associated with each Image so that in Mapper i can figure out the IMAGE ID and then process that image ID based on my given reference file Note: I have one reference file which contains information like Image Number, position co-ordinates of image. So in MapTask when i get any specific image say image number 10 then i do some image processing like drawing shape of given co-ordinates etc
If i use HIPI the map task contains ImageHeader as Key and FloatImage as value .And in the ImageHeader i dont have any information (not even file name) .So i have no way to compare a hdfs distributed files with the reference data i have
Do we have any other better way to do this ?