linklogo Density Image Converter Tool for Android, iOS, Windows and CSS

This is a powerful little tool that helps converting single or batches of images to Android, iOS, Windows and CSS specific formats and density versions given the source scale factor or target width/height in dp. It has a graphical and command line interface and supports a wide array of image types for reading and conversion including PNG, JPEG, SVG, PSD and Android 9-patches. Using sophisticated scaling algorithms, it is designed to make conversion of images easy and fast while keeping the image quality high (comparable to PS). To further optimize the output post processors like pngcrush and mozJpeg can be used (see section below).

GitHub release Build Status Coverage Status Maintainability


1java -jar ./dconvert.jar -src "C:/your-folder/image-folder" -scale 4


1java -jar ./dconvert.jar

will start the UI (or double click the jar file).

To use the post processor (pngcrush, mozJpeg and webp) the respective cmd tool must be set in PATH.


Grab jar from latest release.




Start with

1java -jar ./dconvert.jar

provides the same features as the command line tool so see the help page of command line or tooltips. Additionally the UI will save its settings persistently making it easier for power users.


linkCommand Line

E.g. with xxxhdpi (x4) source files:

1java -jar ./dconvert.jar -src "C:/master-image/ic_my_icon.png" -scale 4 -platform android

Will generate mdpi, hdpi, etc. folders in "C:/master-image/" containing the resized images

Full list of arguments:

1-androidIncludeLdpiTvdpi Android only: creates mipmap sub-folders instead of drawable.

2-androidMipmapInsteadOfDrawable Android only: If set will include additional densities (ldpi and

3 tvdpi)

4-antiAliasing Anti-aliases images creating a little more blurred result; a light

5 3x3 convolve matrix is used; useful for very small images

6-clean Deletes all file and folders in out dir that would be used in

7 current configuration before converting.

8-compressionQuality <0.0-1.0> Only used with compression 'jpg' sets the quality [0-1.0] where 1.0

9 is the highest quality. Default is 0.9

10-dryRun Will not create any images or folder. Useful as fast preview in log

11 what images in what resolutions would be created.

12-dst The directory in which the converted files will be written. Will use

13 the source folder if this argument is omitted.

14-gui Starts graphical user interface

15-h,--help This help page

16-haltOnError If set will stop the process if an error occurred during conversion

17-iosCreateImagesetFolders iOS only: Will create .imageset folders and Content.json for every

18 source image n iOS converter. Default is just all images in the root

19 folder.

20-keepOriginalPostProcessedFiles If a post processor is run on a file, this flag will ensure that the

21 original will be kept (renamed _orig), otherwise only the optimized

22 image will be kept

23-outCompression Sets the compression of the converted images. Can be 'png', 'jpg',

24 'gif', 'bmp', 'png+jpg' or 'strict' which tries to use same

25 compression as source. By default will convert to png except if

26 source compression is jpeg.

27-platform Can be 'all', 'android', 'ios', 'win' or 'web'. Sets what formats

28 the converted images will be generated for. E.g. set 'android' if

29 you only want to convert to android format. Default is [IOS,


31-postProcessorMozJpeg Will post-process all jpegs with mozJpegs lossless compressor

32 'jpegtran'. Mozjpeg (specifically 'jpegtran') must be set in PATH,

33 tested with mozJpeg 3. Binaries for mozJpeg can be found on the

34 internet.

35-postProcessorPngCrush Will post-process all pngs with pngcrush, a lossless compressor. The

36 executable must be set in the system PATH as 'pngcrush' i.e

37 executable from every path. Pngcrush is a tool to compress pngs.

38 Requires v1.7.22+

39-postProcessorWebp Will additionally convert all png/gif to lossless wepb and all jpg

40 to lossy webp with cwebp. Does not delete source files. The

41 executable must be set in the system path as 'cwebp' i.e executable

42 from every path. cwebp is the official converter from Google.

43-roundingMode Defines the rounding mode when scaling the dimensions. Possible

44 options are 'round' (rounds up of >= 0.5), 'floor' (rounds down) and

45 'ceil' (rounds up). Default is ROUND_HALF_UP

46-scale <[float]|[int]dp> The source. Can be an image file or a folder containing image files

47 to be converted. This argument is mandatory.

48-scaleIsHeightDp If set and scale is in dp it will be interpreted as fixed height not

49 width

50-skipExisting If set will not overwrite a already existing file

51-skipUpscaling If set will only scale down, but not up to prevent image quality

52 loss

53-src The source scale. This can either be a factor (1,1.5,2,3,4,etc.)

54 used if the images already have the correct resolution for one scale

55 factor and up- and downscaling for all other densities are needed.

56 Ie. if you have the src file in density xxxhdpi you pass '4'. You

57 could also pass a value in dp (density independent pixels) which

58 denotes the output pixel width (or height if the flag is set) in

59 mdpi/x1. In this mode all output images will have the same width

60 (height). This argument is mandatory.

61-threads <1-8> Sets the count of max parallel threads (more is faster but uses more

62 memory). Possible values are 1-8. Default is 4

63-v,--version Gets current version

64-verbose If set will log to console more verbose

linkUsing the .exe Launcher

Launch4J is used to create Windows executables. Two types are available: dconvert.exe and dconvert-gui.exe. The former is for command line, the later as a simple GUI starter. Launch4J will automatically download the required JRE.


linkScale Factor vs. Fixed DP Width/Height

There are 2 modes available for the user:

linkScale Factor Mode

This mode is based on the idea that a developer gets the 'master' images in the highest needed resolution for his platform, e.g. xxxhdpi (x4) in Android. He then proceeds to generate every other density with it.

To better understand, here is a practical example for Android: Source file is ic_my_icon.png in density xxxhdpi with resolution 144x144 (this will be 36dp x 36dp). The tool will generate the following images in the following resolutions:

1* mdpi 36x36 (x1)

2* hdpi 54x54 (x1.5)

3* xhdpi 72x72 (x2)

4* xxhdpi 108x108 (x3)

5* xxxhdpi 144x144 (x4)

linkFixed Width/Height in DP Mode

This mode is suitable if the source image is in an arbitrary resolution (but usually greater resolution than needed) and the needed density independent pixel (dp) dimensions are known to the developer. Either the width or height can be fixed with a dp value, while calculating the other dimension accordingly (keeping aspect ratio). In this mode all images will have the same pixel width (or height) for the same density bucket. Here is a practical example with fixed width:

You get an icon in 512x512 and want to display it in 48x48dp. The tool will generate the following resolutions:

1* mdpi 48x48 (x1)

2* hdpi 72x72 (x1.5)

3* xhdpi 108x108 (x2)

4* xxhdpi 144x144 (x3)

5* xxxhdpi 192x192 (x4)

linkSupported Platforms

This tool can convert to formats used in Android, iOS and Windows:

There are some options that might change the format of the output. See cmd help or tooltips for more info.

linkSupported File Types

This tool can read & write the following file types:

Details on the supported features of each format see here.

Out-compression has the following modes:

If an image will be re-compressed with a file type that does not support alpha, alpha will be replaced with white (e.g. when converting png -> jpeg)

linkQuality Comparison

Note this will be available in future versions.

One of the main features of this converter is downscaling. Unfortunately without using specialized algorithms, down scaled images may look jagged or too blurred. I tested the algorithms of Thumbnailator and imgscalr which basically uses Chris Campbell's proposed progressive scaling algorithm. Upon further analysis Lanczos3 filtering seems to be recommended which produces results preserving more detail without being too jagged. The downside is, it is considerably slower than the former. This tool uses the implementation of Morten Nobel.

The tool provides 3 quality profiles:

In most cases Lanczos3 creates best results, although in some cases progressive bilinear scaling may be superior for very sharp edges of e.g. simple icons where the former creates some softness. Here are some example to see the difference yourself. The icons are down-scaled from 96x96 to 72x72 and 48x48, the photos from different source from 500 to 2000 px.

comparison comparison

comparison comparison

As reference, here are the results of other methods:


  1. Thumbnailator with default settings (progressive bilinear)
  2. Photoshop CS5 with bicubic algorithm
  3. imgscalr with ULTRA_QUALITY setting (progressive bilinear)

linkPost Processors

The converter tool supports some post processors which run after the main conversation task to manipulate the resulting files. Post processors can be easily implemented by implementing the PostProcessor interface.


Pngcrush is a brute force png lossless compression tool. This may be already in your build chain (Android does this automatically), but may be useful if not. Will run pngcrush over the all converted *.png files. See code or output for specific used options.

The converter tool requires pngcrush to be in PATH. Minimal required version is 1.7.22; currently tested with v1.7.87 & 1.8.0. More info on the pngcrush wesbite


WebP is an image format employing both lossy and lossless compression developed by Google. It is natively supported in Android and claims to be 25-35% smaller in file size than png and jpeg (in their respective categories).

If the flag is set the converter tool will compress jpegs to lossy webp and png/gif to lossless version of it. The program cwebp, which can be downloaded from Google's page, must be set in PATH variable.

More info on webp


MozJpeg is a Jpeg encoder that also features some lossless compression techniques developed by the Mozilla foundation. It can be used to further shrink jpegs in size. Unfortunately they do not provide binaries on github, but they can be found e.g. here. The tool requires jpegtran to be set in PATH. Tested with mozJpeg 3.1

More info on mozJpeg

linkImplementation Details

Java 11, JavaFx and Apache CLI

For support of various image types TwelveMonkey lib is used which claims to have wider support of even basic file types (jpeg, etc.)

For parallel execution, every source file spawns his own thread for each platform converter e.g. 3 files 2 converters -> 6 threads needed.

The tool can easily be extended by using the IPlatformConvert and PostProcessor interfaces.

Both user interfaces use the same underlying logic.

linkLimitations & Known Issues

linkPossible Future Features

linkHow to build

Use the Maven wrapper to create a jar including all dependencies

1mvnw clean package

The .jar file will be in the generated /target folder

linkCheckstyle Config File

This project uses my common-parent which centralized a lot of the plugin versions aswell as providing the checkstyle config rules. Specifically they are maintained in checkstyle-config. Locally the files will be copied after you mvnw install into your target folder and is called target/checkstyle-checker.xml. So if you use a plugin for your IDE, use this file as your local configuration.




Copyright 2016 Patrick Favre-Bulle

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at


Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

Density Image Converter Tool for Android, iOS, Windows and CSSDownloadRequirementsUsageGUICommand LineUsing the .exe LauncherDetailsScale Factor vs. Fixed DP Width/HeightScale Factor ModeFixed Width/Height in DP ModeSupported PlatformsSupported File TypesQuality ComparisonPost ProcessorspngcrushwebpmozJpegImplementation DetailsLimitations & Known IssuesPossible Future FeaturesHow to buildCheckstyle Config FileAboutCreditsLicense



Patrick Favre-Bulle 2020