python - How to check is Christmas tree image edited -
we 3 friends between started funny competition find best method programmatically check in image christmas tree toy edited or not.
i decided more unique , after long search have chosen error level analysis. method show edited places more brighter , identified anomaly easier human eyes identify programmatically have mission:)
i decided use python , opencv i'm newbie think can done objective.
i'm still not sure method use find anomaly in image after error level analysis. opinion easiest way find brightest object in image, practically found not work in cases, find other regions brighter or find brightest region in original, not edited image , return image edited.
in cases after editing object in image object become more brighter in cases can become specific color, noisier, easier recognized human eyes.
what ideas have:
- firstly, need find christmas tree toy biggest, main object in image, i'm thinking add thresholding , find biggest white object , hope christmas tree toy.
- 1.1) convert image pixels blocks, compute each block brightness average formula
(.299 * red) + (.587 * green) + (.114 * blue)
, select brightest blocks , check in christmas toy region. if return christmas tree toy edited. - 1.2) compute whole image pixels brightness average same formula, compute christmas tree toy region pixels brightness average. compare whole image brightness christmas tree brightness if toy region brighter when return image edited.
but afraid methods not work how expect. whole image brightness or other image regions in situations can brighter edited region. not have many ideas experts appreciated.
original image (toy angles can various, biggest object in image):
original image after error level analysis:
first edited image (changed color, added noise):
first edited image after error level analysis:
second edited image (changed color, added gaussian blur , text "happy christmas!"):
second edited image after error level analysis:
Comments
Post a Comment