Multiscale gradation processing method

Abstract

A method of generating a contrast enhanced version of a grey value image by applying contrast amplification to a multiscale representation of the grey value image wherein density in the contrast enhanced version as a function of grey value and contrast amplification are specified independently.

1 . A method of generating a contrast enhanced version of a grey value image by applying contrast amplification to a multiscale representation of said image, 
 said contrast enhanced version being obtained by applying a reconstruction process to said multi-scale representation whereby a scale-specific conversion function is inserted at each successive stage of said reconstruction process from a predefined large scale on so that the output of a stage of said reconstruction process is converted by a conversion function specified for that scale before being supplied to the input of a next stage of the reconstruction process    wherein    a specification of said contrast amplification as a function of grey value at two or more successive scales is defined in advance and the conversion functions for each of said successive scales are derived from said specifications.    
     
     
         2 . A method according to  claim 1  wherein the scale-specific conversion functions are derived from a series of scale-specific gradient functions that specify the amount of contrast amplification as a function of grey value at successive scales.  
     
     
         3 . Method according to  claim 2  wherein a gradient function for a predefined large scale among said scales is the derivative of a predefined gradation function that specifies density as a function of grey value.  
     
     
         4 . A method according to  claim 3  wherein said large-scale gradation function has a predefined ordinate value and a predefined slope in an anchor point, the abscissa value of the anchor point being deduced from a digital image representation of said grey value image or of a large-scale image obtained by applying partial reconstruction to said multiscale representation.  
     
     
         5 . A method according to  claim 3  wherein said large-scale gradation function has a predefined shape, and is stretched and shifted along the abscissa axis in order to match a relevant subrange of pixel values of said grey value image or of a large-scale image obtained by applying partial reconstruction to said multiscale representation.  
     
     
         6 . Method according to  claim 2  wherein a gradient function for a predefined large scale among said scales is derived from the histogram of the pixel values of said grey value image or from the histogram of pixel values of a large scale image obtained by applying partial reconstruction to said multiscale representation.  
     
     
         7 . A method according to  claim 3  wherein said large-scale gradation function is derived from the histogram of pixel values of said grey value image or of a large-scale image obtained by applying partial reconstruction to said multiscale representation.  
     
     
         8 . A method according to  claim 7  in which said large-scale gradation function is further adjusted so that it has a predefined ordinate value in at least one anchor point, the abscissa of which is determined as a characteristic point of the histogram of pixel values of said grey value image or of a large-scale image obtained by applying partial reconstruction to said multiscale representation.  
     
     
         9 . A method according to  claim 2  wherein a gradient function for a predefined small scale is predefined.  
     
     
         10 . A method according to  claim 9  wherein said gradient function at said predefined small scale has a predefined value in each of at least two overlapping grey value bands.  
     
     
         11 . A method according to  claim 9  wherein said predefined gradient function for said small scale is expressed as a function of density.  
     
     
         12 . A method according to  claim 10  wherein said predefined gradient function for said small scale is expressed as a function of density.  
     
     
         13 . A method according to  claim 9 , modified so that the small-scale gradient function is adjusted as a function of the signal-to-noise ratio of the original digital image.  
     
     
         14 . A method according to  claim 10 , modified so that the small-scale gradient function is adjusted as a function of the signal-to-noise ratio of the original digital image.  
     
     
         15 . A method according to  claim 11 , modified so that the small-scale gradient function is adjusted as a function of the signal-to-noise ratio of the original digital image.  
     
     
         16 . A method according to  9 , wherein gradient functions at the scales smaller than said small scale, are identical to the gradient function for said small scale.  
     
     
         17 . A method according to  claim 2 , in which gradient functions for intermediate scales in between said large scale and a predefined small scale have a shape that evolves, gradually from the shape of the gradient function for said large scale to the shape of a gradient function for said predefined small scale.  
     
     
         18 . A method according to  claim 17 , in which the gradient functions gm k () at intermediate scales k are defined by:  
       
         
           
             
               
                 
                   
                     gm 
                     k 
                   
                   ( 
                   ) 
                 
                 = 
                 
                   
                     
                       gm 
                       S 
                     
                     ( 
                     ) 
                   
                   · 
                   
                     
                       ( 
                       
                         
                           
                             gm 
                             L 
                           
                           ( 
                           ) 
                         
                         
                           
                             gm 
                             S 
                           
                           ( 
                           ) 
                         
                       
                       ) 
                     
                     
                       
                         k 
                         - 
                         S 
                       
                       
                         L 
                         - 
                         S 
                       
                     
                   
                 
               
               , 
             
           
           
           
               
           
         
       
       where gm L () is said large-scale gradient function at scale L, gm S () is said small-scale gradient function at scale S, and S<k<L.  
     
     
         19 . A method according to  claim 2  in which one or more of the scale-specific gradient functions or scale-specific conversion functions are stored as lookup tables.  
     
     
         20 . A method according to  claim 1 , in which said multiscale representation is a Burt pyramid, a multiresolution subband representation or a wavelet representation.  
     
     
         21 . A method according to  claim 1 , in which said grey value image is a medical image.  
     
     
         22 . A method according to  claim 21 , in which said medical image is a digital X-ray image.  
     
     
         23 . A computer program product adapted to carry out the method of  claim 1  when run on a computer.  
     
     
         24 . A computer readable medium comprising computer executable program code adapted to carry out the steps of  claim 1.