accumarray error: Second input VAL must be a vector with one element for each row in SUBS, or a scalar

 Error using accumarray
Second input VAL must be a vector with one element for each row in SUBS, or a scalar.

Error in descriptFPoints (line 160) DescriptVector = accumarray( AngleBinsShifted, SubPatch_norm(IndexMaskOfBlocks), [ NOfBins*NOfWindows^2, 1 ] ); My code segment is given below:

    for i = 1:N

            %---------------- Data of feature point ---------------%
            Scale = HrLOrntPoints(i,3);
            MainOrient = deg2rad(HrLOrntPoints(i,6));
            row0 = HrLOrntPoints(i,4);
            clmn0 = HrLOrntPoints(i,5);
            Step = StepSampleFunction( Scale );

            %-------------- Neighborhood in new coord -------------%
            % new coordinates received from old ones by rotation, and shift by [ row0, clmn0 ]
            mx  = max( ceil(2*NOfWindows*Step), round(3*Scale) + 1 );
            ind = -mx:mx;
            % pay attention, that coordinates are taken in column-major order, so
            % assignments below of Patch is consistent
            [ Clmn, Row ] = meshgrid( ind, ind );
            % tmp2 need to be >= 4*NOfWindows, in order to descript point after
            % mx, that was choosen before, provide this
            tmp2 = 2*mx + 1; tmp = tmp2^2;
            Coord_new = zeros( tmp, 2 );
            Coord_new( :, 2 ) = Row(:);
            Coord_new( :, 1 ) = Clmn(:);
            Coord_new = Coord_new';

            %------ Converting coordinates to original form -------%
            Rotation = [ cos(MainOrient) -sin(MainOrient); sin(MainOrient) cos(MainOrient) ];
            % Coord_orig(1,:) - column coord, that correspond to Coord_new( 1, : ).
            % Coord_orig(2,:) - rows ...
            Coord_orig = round( (Rotation)*Coord_new + repmat( [ clmn0 row0 ]', 1, tmp ) );

            %---------- Converting to linear indexes  -------------%
            IndOfRotatedPatch = sub2ind( [ Nrow, Nclmn ], Coord_orig( 2,: )', Coord_orig( 1,: )' );

            %---------- Assigning values to the patch  -------------%
            Patch = zeros( tmp2 );
            Patch(:) = img( IndOfRotatedPatch );

            %------------------ Derivative mask -------------------%
            ind = -round(3*Scale):round(3*Scale);
            [ X, Y ] = meshgrid( ind, ind );
            dGdx = -X .* exp(-( X.*X + Y.*Y )/(2*Scale*Scale)) ./ ((Scale^4)*(2*pi));
            dGdy = -Y .* exp(-( X.*X + Y.*Y )/(2*Scale*Scale)) ./ ((Scale^4)*(2*pi));

            %----------------- Window derivatives -----------------%
            % also possible to do convolution, cause gaussian symmetric
            Patchx = imfilter(Patch, dGdx, 'same');
            Patchy = imfilter(Patch, dGdy, 'same');

            %----------------- Norm of gradients ------------------%
            gradient_norm = sqrt(Patchx.^2 + Patchy.^2);

            %------------- Avaraging norm gradients --------------%
            g   = fspecial('gaussian',max(1,fix( 6*Scale*Factor )), Scale*Factor);
            gradient_norm = imfilter(gradient_norm, g,  'same');

            %----------------- Angles of gradients ----------------%
            gradient_angles = rad2deg( atan2( Patchy, Patchx ) + pi );

            %--------- Taking sub-window for description ----------%
            % I'm taking 'center pixel'( Feature point, with coordinates [ row0, clmn0 ] in image ) to be in
            % new/sub-patch coordinates at [ p, p ]
            % p = round(tmp2/2)
            p = mx + 1;
            % so I'll have patches of size 4*NOfWindows x 4*NOfWindows
            SubPatch_norm  = gradient_norm( max(round( (p  - ( 2*NOfWindows )*Step):Step:(p  + ( 2*NOfWindows -1  )*Step) ), 1), max(round( (p  - ( 2*NOfWindows )*Step):Step:(p + ( 2*NOfWindows -1  )*Step)), 1) );
            SubPatch_angles  = gradient_angles( max(round( (p  - ( 2*NOfWindows )*Step):Step:(p  + ( 2*NOfWindows -1  )*Step) ), 1), max(round( (p - ( 2*NOfWindows )*Step):Step:(p + ( 2*NOfWindows -1  )*Step)), 1) );

            %-------------- Assigning angles 2 bins ---------------%
            [tmp, AngleBins] = histc( SubPatch_angles(IndexMaskOfBlocks), Edges4Hist );

            %---------------- Shifting angle bins -----------------%
            AngleBinsShifted = AngleBins + IndexShiftBins4Accum;

            %-------- Accumulating norms 2 according bins ---------%
            DescriptVector = accumarray(AngleBinsShifted, SubPatch_norm(IndexMaskOfBlocks), [ NOfBins*NOfWindows^2, 1 ] );
            DescriptVector = min( DescriptVector/norm(DescriptVector), Thresh );
            DescriptVector = DescriptVector/norm(DescriptVector);

            %---------------- Assign values 2 matrix --------------%
            DescriptFPVector( i, : ) = (DescriptVector)';
            %     if strcmp( SwitchWaitbars, 'on' )
            %     waitbar(i/N)
            %     end
        end

[Jan]

We cannot run your code. The error message seems to be clear already. So how exactly can we help you?

Set a breakpoint in this line and check the sizes of the vectors. Maybe a transposing is enough already.

[K M Ibrahim Khalilullah]

@Jan Simon Thank you very much for your comment. Actually I downloaded the code from Matlab file exchange. Would you please check it from this link? https://uk.mathworks.com/matlabcentral/fileexchange/29004-feature-points-in-image--keypoint-extraction It is not working for all images. My image is greater than 5 MB, so it is not possible to upload.

Answer 1

[K M Ibrahim Khalilullah]

Solved this problem by avoiding third argument of the function. The program is Ok now. Thank to all