% CYCLEMIX Multi-image blending over a cyclic sequence of images % % Usage: cyclemix(im, figNo, nodeLabel) % % Arguments: % im - A cell array of images to be blended. If omitted, or % empty, the user is prompted to select images via a file % dialog. % figNo - Optional figure number. % nodeLabel - Optional cell array of strings specifying the labels to % be associated with the nodes on the blending interface. % % This function sets up an interface consisting of a circle with equispaced % nodes around it corresponding to the input images to be blended. An % additional virtual node corresponding to the average of all the images is % placed at the centre of the circle. Positioning the mouse at some location % within the circle generates an interactive blend formed from the weighted sum % of the two image nodes that the mouse is between (in an angular sense) and % between the central average image according to the radial position of the % mouse. % % An application where this blending tool can be useful is to blend between a % sequence of images that have been filtered according to a parameter that is % cyclic, say an orientation filter. % % See also: LINIMIX, BILINIMIX, CLIQUEMIX, TERNARYMIX % Reference: % Peter Kovesi, Eun-Jung Holden and Jason Wong, 2014. % "Interactive Multi-Image Blending for Visualization and Interpretation", % Computers & Geosciences 72 (2014) 147-155. % http://doi.org/10.1016/j.cageo.2014.07.010 % Copyright (c) 2011-2014 Peter Kovesi % Centre for Exploration Targeting % The University of Western Australia % peter.kovesi at uwa edu au % % Permission is hereby granted, free of charge, to any person obtaining a copy % of this software and associated documentation files (the "Software"), to deal % in the Software without restriction, subject to the following conditions: % % The above copyright notice and this permission notice shall be included in % all copies or substantial portions of the Software. % % The Software is provided "as is", without warranty of any kind. % August 2011 Original version % March 2014 Cleanup and documentation function cyclemix(im, figNo) if ~exist('im', 'var'), im = []; end if ~exist('figNo', 'var') || isempty(figNo) fig = figure; else fig = figure(figNo); end [im, nImages, fname] = collectncheckimages(im); % Compute average image avim = zeros(size(im{1})); for n = 1:nImages avim = avim + im{n}; end avim = avim/nImages; % Generate vertices of blending polygon interface deltaTheta = 2*pi/nImages; theta = [0:(nImages-1)] * deltaTheta; v = [cos(theta) sin(theta)]; % Display 1st image and get handle to Cdata in the figure % Suppress display warnings for images that are large S = warning('off'); figure(fig), clf, imposition = [0.35 0.0 0.65 1.0]; subplot('position', imposition); imshow(normalise(im{1}), 'border', 'tight'); set(fig, 'NumberTitle', 'off') set(fig, 'name', ' CycleMix') set(fig, 'Menubar','none'); imHandle = get(gca,'Children'); warning(S); % Set up interface figure ah = subplot('position',[0.0 0.4 0.35 0.35]); h = circle([0 0], 1, 36, [0 0 1]); set(h,'linewidth',2) hold on plot(v(1,:), v(2,:), '.', 'color', [0 0 0], 'markersize', 40') plot(v(1,:), v(2,:), '.', 'color',[.9 .9 .9], 'markersize',20) radsc = 1.2; for n = 1:nImages if v(1,n) < -0.1 text(v(1,n)*radsc, v(2, n)*radsc, namenpath(fname{n}), ... 'FontSize', 16, 'FontWeight', 'bold',... 'HorizontalAlignment','right'); elseif v(1,n) > 0.1 text(v(1,n)*radsc, v(2, n)*radsc, namenpath(fname{n}), ... 'FontSize', 16, 'FontWeight', 'bold',... 'HorizontalAlignment','left'); else text(v(1,n)*radsc, v(2, n)*radsc, namenpath(fname{n}), ... 'FontSize', 16, 'FontWeight', 'bold',... 'HorizontalAlignment','center'); end end r = 1.25; axis ([-r r -r r]), axis off, axis equal % Set callback function set(fig,'WindowButtonDownFcn',@wbdcb); [rows, cols, ~] = size(im{1}); setwindowsize(fig, [rows cols], imposition); % Set up custom pointer myPointer = [circularstruct(7) zeros(15,1); zeros(1, 16)]; myPointer(~myPointer) = NaN; fprintf('\nLeft-click to toggle blending on and off.\n'); blending = 0; hspot = 0; %---------------------------------------------------------------- % Window button down callback function wbdcb(src,evnt) if strcmp(get(src,'SelectionType'),'normal') if ~blending % Turn blending on blending = 1; set(src,'Pointer','custom', 'PointerShapeCData', myPointer,... 'PointerShapeHotSpot',[9 9]) set(src,'WindowButtonMotionFcn',@wbmcb) if hspot % For paper illustration delete(hspot); end else % Turn blending off blending = 0; set(src,'Pointer','arrow') set(src,'WindowButtonMotionFcn','') % For paper illustration cp = get(ah,'CurrentPoint'); x = cp(1,1); y = cp(1,2); if gca == ah hspot = plot(x, y, '.', 'color', [0 0 0], 'Markersize', 40'); end end end end %---------------------------------------------------------------- % Window button move call back function wbmcb(src,evnt) cp = get(ah,'CurrentPoint'); x = cp(1,1); y = cp(1,2); blend(x,y) end %------------------------------------------------------------------- function blend(x,y) % Get radius and angle of cursor relative to centre of circle radius = min(norm([x y]), .999); % Clamp radius to a max of 1 ang = atan2(y,x); % Define a sigmoidal weighting function for radius values ranging from 0 to % 1 yielding weights that range from 1 to 0. This is the weighting that is % applied to the average image associated with the centre of the circular % interface. Using a sigmoidal function provides a slight 'flat spot' at % the centre (and edges) of the interface where the weights do not change % much. This makes the interface a bit easier to use near the centre % which otherwise forms a singularity with respect to cursor movements. radialWeight = logisticweighting(radius, 2^4, [0 1 1 0]); % Compute angular distance to all vertices ds = sin(theta) * cos(ang) - cos(theta) * sin(ang); % Difference in sine. dc = cos(theta) * cos(ang) + sin(theta) * sin(ang); % Difference in cosine. distTheta = abs(atan2(ds,dc)); % Absolute angular distance. % First form a 2-image blend from the two images that are on each side of % the cursor in an angular sense % Zero out angular distances > deltaTheta, these will correspond to nodes % that are not on each side of the cursor in an angular sense. There should % be only two non-zero elements after this. distTheta(distTheta > deltaTheta) = 0; % Form normalised angular blending weights. angularWeight = (deltaTheta-distTheta)/deltaTheta; % Identify the indices of the non-zero weights ind = find(distTheta); if length(ind) == 2 blend = angularWeight(ind(1))*im{ind(1)} + angularWeight(ind(2))*im{ind(2)}; elseif length(ind) == 1 % Unlikely, but possible. blend = angularWeight(ind(1))*im{ind(1)}; end % Now form a 2-image blend between this image and the centre, average image. blend = (1-radialWeight)*blend + radialWeight*avim; blend = normalise(blend); set(imHandle,'CData', blend); end % end of blend %--------------------------------------------------------------------- end % of main function and nested functions %------------------------------------------------------------------------ % Function to set up window size nicely function setwindowsize(fig, imsze, imposition) screen = get(0,'ScreenSize'); scsze = [screen(4) screen(3)]; window = get(fig, 'Position'); % Set window height to match image height window(4) = imsze(1); % Set window width to match image width allowing for fractional width of % image specified in imposition window(3) = imsze(2)/imposition(3); % Check size of window relative to screen. If larger rescale the window % size to fit 80% of screen dimension. winsze = [window(4) window(3)]; ratio = max(winsze./scsze); if ratio > 1 winsze = winsze/ratio * 0.8; end window(3:4) = [winsze(2) winsze(1)]; set(fig, 'Position', window); end