ImageResize4J - Tutorial

TUTORIAL

Introduction

ImageResize4J offers 24 interpolation filters for high-quality image resize and scaling. Interpolation filter selection depends on a current task: the fastest interpolation filter is the NEAREST_NEIGHBOR and the most qualitative are IDEAL filter families and SHARP (SHARP_LIGHT/SHARP_MORE) filter families. Other interpolation filters are the balance between speed & quality.

! SHARP (SHARP_LIGHT/SHARP_MORE) filter families are the most appropriate solution for producing sharp and nice-looking photo thumbnails.

! Interpolation filters with more length are more qualitative but slower.

Java2D-based Image Scaling


        //source image

        BufferedImage sourceImage = ...;

        

        //Create an instance of ImageResizeProcessor class

        //with the "Sharp Light" interpolation filter of length=5

        ImageResizeProcessor processor =

                new ImageResizeProcessor(ImageResizeProcessor.TYPE_SHARP_LIGHT_5);

        

        //scale BufferedImage by factor = 0.2 with maintaining the aspect ratio

        //of the original image dimension

        BufferedImage scaledImage_1 = processor.scale(sourceImage, 0.2f, -1);

        

        //scale BufferedImage by X-axis scale factor = 0.8 

        //and Y-axis scale factor = 0.6

        BufferedImage scaledImage_2 = processor.scale(sourceImage, 0.8f, 0.6f);

TOP

Java2D-based Image Resize


        //source image

        BufferedImage sourceImage = ...;

        

        //Create an instance of ImageResizeProcessor class

        //with the "Sharp Light" interpolation filter of length=5

        ImageResizeProcessor processor =

                new ImageResizeProcessor(ImageResizeProcessor.TYPE_SHARP_LIGHT_5);

        

        //resize BufferedImage to the width=300px with maintaining

        //the aspect ratio of the original image dimension

        BufferedImage resizedImage_1 = processor.resize(sourceImage, 300, -1);

        

        //resize BufferedImage to the width=300px and height=200px

        BufferedImage resizedImage_2 = processor.resize(sourceImage, 300, 200);

TOP

JAI-based Image Scaling


        //source image

        PlanarImage sourceImage = ...;

        

        //scale PlanarImage by factor 0.2 with maintaining the aspect ratio

        //of the original image dimension

        //using "Sharp Light" interpolation filter of length=5

        RenderedOp renderedOp_1 =

                ImprovedScaleDescriptor.create(sourceImage,

                        new Float(0.2f), null,

                        ImprovedScaleInterpolation.SHARP_LIGHT_5, null);

        

        //scale PlanarImage by X-axis scale factor = 0.8

        //and Y-axis scale factor = 0.6

        //using "Sharp Light" interpolation filter of length=5

        RenderedOp renderedOp_2 =

                ImprovedScaleDescriptor.create(sourceImage,

                        new Float(0.8f), new Float(0.6f),

                        ImprovedScaleInterpolation.SHARP_LIGHT_5, null);

TOP

JAI-based Image Resize


        //source image

        PlanarImage sourceImage = ...;

                

        //resize PlanarImage to the width=300px with maintaining

        //the aspect ratio of the original image dimension

        //using "Sharp Light" interpolation filter of length=5

        RenderedOp renderedOp_1 =

                ImprovedScaleDescriptor.createResize(sourceImage,

                        new Integer(300), null,

                        ImprovedScaleInterpolation.SHARP_LIGHT_5, null);

                

        //resize PlanarImage to the width=300px and height=200px

        //using "Sharp Light" interpolation filter of length=5

        RenderedOp renderedOp_2 =

                ImprovedScaleDescriptor.createResize(sourceImage,

                        new Integer(300), new Integer(200),

                        ImprovedScaleInterpolation.SHARP_LIGHT_5, null);

TOP

Performance Tips

To produce high-quality thumbnails from large images with an appropriate speed s.c. "two phase" approach may be used:

First Phase generates an intermediate image using the faster interpolation filter likes NEAREST_NEIGHBOR;
Second Phase generates one-or-more high quality thumbnails using the more qualitative interpolation filters like IDEAL or SHARP (SHARP_LIGHT/SHARP_MORE) families;

The below results demonstrate the advantage of such approach (example source code, source image):

1-Phase Java2D Bicubic	1-Phase IR4J Nearest Neighbor	1-Phase IR4J Sharp 5	2-Phase IR4J Nearest Neighbor/Sharp 5
72x96 thumbnails
Average processing: 1014ms	Average processing: 128ms	Average processing: 1140ms	Average processing: 139ms
96x128 thumbnails
Average processing: 1012ms	Average processing: 121ms	Average processing: 1107ms	Average processing: 169ms
120x160 thumbnails
Average processing: 1114ms	Average processing: 119ms	Average processing: 1014ms	Average processing: 202ms