Reconstructing an image of a scene captured using a compressed sensing device

A system for reconstructing an image of a scene captured using a compressed sensing device, the system comprising:

• the compressed sensing device comprising a spatial light modulator configuring a plurality of spatial patterns according to a set of pseudorandom basis functions, such that said basis functions exist on more than one lattice, each having spatially varying spatial resolution, a spatial resolution of said pseudorandom basis functions being adaptively modified according to attributes of a mask which identifies at least one region of interest in an image of a scene, each pattern of said spatial light modulator focuses a portion of the incoming light of said scene onto at least one detector which samples sequential measurements of light focused thereon by said pattern, each of said measurements comprising an inner product result ym=<x,φm>, where x denotes a N-dimension vector representing a N-pixel sampled version of said scene and φm denotes a mth spatially basis function used for sampling, a series of said measurements comprising a sequence of projection coefficients corresponding to said scene;
• and
• a processor executing machine readable program instructions for performing the method of:
• receiving linear projection measurements of said scene obtained using the compressed sensing device;
• and
• reconstructing an appearance of said scene utilizing a compressed sensing framework which reconstructs said image from said sequence of projection coefficients after M inner products have been sampled, where M is smaller than N, said reconstruction being such that pixel locations associated with said identified at least one region of interest have a higher image quality spatial resolution when rendered relative to other pixels of said image, such that a resolution of a lattice will be larger on said pixel locations associated with said identified at least one region of interest.

A computer implemented method for reconstructing an image of a scene captured using a compressed sensing device, the method comprising:

• obtaining linear projection measurements of a scene using the compressed sensing device which comprises a spatial light modulator configuring a plurality of spatial patterns according to a set of pseudorandom basis functions, such that said pseudorandom basis functions exist on more than one lattice, each having spatially varying spatial resolution, a spatial resolution of said pseudorandom basis functions being adaptively modified according to attributes of a mask which identifies at least one region of interest in an image of said scene, each pattern of said spatial light modulator focuses a portion of the incoming light of said scene onto at least one detector which samples sequential measurements of light focused thereon by said pattern, each of said measurements comprising an inner product result ym=<x,φm>, where x denotes a N-dimension vector representing a N-pixel sampled version of said scene and φm denotes a mth spatially incoherent basis function used for sampling, a series of said measurements comprising a sequence of projection coefficients corresponding to said scene;
• and
• reconstructing an appearance of said scene utilizing a compressed sensing framework which reconstructs said image from said sequence of projection coefficients after M inner products have been sampled, where M is smaller than N, said reconstruction being such that pixel locations associated with said identified at least one region of interest have a higher image quality spatial resolution when rendered relative to other pixels of said image, such that a resolution of a lattice will be larger on said pixel locations associated with said identified at least one region of interest.