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).
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).
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.
xxxhdpi (x4) source files:
1java -jar ./dconvert.jar -src "C:/master-image/ic_my_icon.png" -scale 4 -platform android
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
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.
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
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
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.
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
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.
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
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
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
Launch4J is used to create Windows executables. Two types are available:
dconvert-gui.exe. The former is for command line, the later
as a simple GUI starter. Launch4J will automatically download the required JRE.
There are 2 modes available for the user:
This mode is based on the idea that a developer gets the 'master' images in the highest needed resolution for his platform,
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)
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)
This tool can convert to formats used in Android, iOS and Windows:
drawable-mdpi, etc. folders and saves images with the original name and respective scale factor in there. 9-Patches are specifically supported for Android - must have
.9.png extension and only size correctly if out-compression is PNG.
.imageset postfix and 3 images in them with no,
@3x postfix also including is the
Content.json; or creates a single folder
AssetCatalog with all the images in it.
Assets folder containing all the images with
.scale-140 etc. postfixes. This should be compatible with Windows universal apps.
img folder containing all the images with a
1x and a
2x scale factor which is compatible with the css image-set definition for multiple densities.
There are some options that might change the format of the output. See cmd help or tooltips for more info.
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)
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:
Lanczos3 filtering (8 times slower than Balanced)
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.
As reference, here are the results of other methods:
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
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.
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
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
Both user interfaces use the same underlying logic.
Use the Maven wrapper to create a jar including all dependencies
1mvnw clean package
The .jar file will be in the generated
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.Star
Overview How to Centralize your Checkstyle Configuration with Maven A Better Way to Protect Your IDs Security Best Practices: Symmetric Encryption with AES in Java and Android: Part 2: AES-CBC + HMAC The Bcrypt Protocol… is kind of a mess The Concise Interface Implementation Pattern Improving ProGuard Name Obfuscation Handling Proguard as Library Developer Managing Logging in a Multi-Module Android App Security Best Practices: Symmetric Encryption with AES in Java and Android
Patrick Favre-Bulle 2020