Estimating jpeg file size

Find centralized, trusted content and collaborate around the technologies you use most. Connect and share knowledge within a single location that is structured and easy to search.

I have to say the I don't know much about how file formats work. I think that the reasoning that led to the question will help some one answer me. I have a Java Applet the uploads Images that people draw with it to my server. I need to know what the max size that this file can conceivably reach. It is always going to be x It sounds dumb but are there colors that take more byte size then others and if so what is the most expensive one?

At the extreme end of the spectrum there is no limit to the size, since the standard doesn't limit some types of marker appearing more than once - e. Huffman symbols can range from 1 to 16 bits and are chosen by the encoder to be as small as possible; However, the choice of symbol length can be specified. Final encoding of the huffman bitstream must be done so that markers can be uniquely identified.

Using all this information we can construct a pathological, but valid, JPEG file which libjpeg the most common JPEG decoder implementation is happy to decode. First, we need the longest possible huffman symbols. At first thought, defining a maximal-length huffman symbol 16 bits of all 1's, would use most space, however libjpeg refuses to handle a Huffman symbol which is all 1's, this doesn't seem to be excluded by the standard - as it still a unique symbol as the size is already known to be 16 bits unlike other variable-length symbols, and indeed some decoders can handle it JPEGSnoop.

Most of these bytes will be 0xff, which are replaced by 0xff,0x00 in the output stream, as per the standard, in order to differentiate the bitstream from valid markers.

Perform this mapping and a complete DCT is represented by 8 repeats of the following byte pattern :. As a rule of thumb, no JPEG is going to be larger than an equivalent size bit bitmap. JPEG converts images to a different color space YCbCr which uses fewer bits 12 to be exact to represent a single color. Realistically speaking though, the image will be much smaller than the above formula would suggest.

If we use lossless compression only, the file size would be a bit smaller. Even then, no one does that so your image should be far below the limit set by that formula. I'm not sure this would be that helpful, but I believe that the absolute maximum it could be would be:. The final size in bytes is based on the encoding quality settings used and the number of pixels. In your case, all images should be the same size since you are doing the encoding and your user seems forced to draw on a x area.

Stack Overflow for Teams — Collaborate and share knowledge with a private group. This multiplication step is also why the quantization tables used to encode the image must be included with the JPEG file in order to decode the image. Decoding with the wrong quantization tables will result in a blurry picture.

Smaller quantization table values means less data loss and a higher quality image. Bigger values result in a much lower quality image. With quantization tables, you will usually see the lower frequencies top-left corner using smaller numbers than the higher frequencies bottom-right corner.

The values used for the quantization tables are specified by the encoder. And an extremely low quality JPEG e. Mapping Values Although the quantization table format and usage is well defined, there is no widely accepted method for translating table values into a descriptive JPEG quality.

For example, Adobe Photoshop offers multiple scaling methods: Save As. One method used by Photoshop is seen when using "Save As" and allows the user to select one of 12 quality levels with names like "Maximum", "High", and "Low". Save for Web.

Another method appears when using "Save for Web" and permits the user to select a quality value from 0 to Photoshop includes an advanced option to save using the JPEG Standard algorithm rather than its own quantization tables. However, this option is buried in the menus and the location varies by software version. A value without a percent sign e. Non-standard quantization tables may also be associated with a text name, such as "High" or "Medium".

Cameras Most digital cameras have different quality settings. These usually have names like "High" and "Low" but do not identify a numerical quality value. Each of these built-in camera quality levels identifies hard-coded quantization tables that are defined in the camera's firmware. It is common for different cameras to use different hard-coded quantization tables with the same textual names.

The "High" setting on a Canon camera is unlikely the same as "High" on an Olympus, Nikon, or other type of camera. The hard-coded quantization tables are often distinct to a specific make and model. Different cameras in the same product line frequently use different different quantization tables.

Instead, it needs to be either identified or estimated. There are a number of approaches for identifying the quality of a JPEG. The quantization table selection directly impacts the JPEG quality. If the quantization tables can be identified as coming from a known device or application, then the quality level is known.

A brute-force approach collects every set of quantization tables from every known camera and application, and then performs a lookup to match the set of quantization tables from a new JPEG image to a known quality level. In rare cases, the application creating the JPEG may store the quality level as a metadata field.

With some Adobe products, there may be an Adobe metadata field that identifies either the quality level or range used for the image. However, the presence and values in these fields vary by Adobe product, version, and even saving options. Metadata is an unreliable source for identifying the quality level. For example, ImageMagick will retain comments and metadata fields generated by other applications.

In general, metadata descriptions of the JPEG quality level are rarely available, inconsistent between applications, and unreliable when present. However, any identified inconsistencies can be used to identify likely resaves. Approximate Ratios. The ratio of values between the JPEG quantization tables can be used to estimate the overall quality level.

Approximate Quantization Tables. If the image was created using the JPEG Standard's scaling method, then this will produce a very accurate result.

For Adobe "Save As", this approach will estimate numerical qualities. Estimation Method FotoForensics uses the Approximate Quantization Tables approach and ignores metadata when estimating the last saved quality.

FotoForensics also indicates whether the quantization tables match the JPEG Standard, or whether they are an approximation from a non-standard set of quantization tables. Because the metadata is not used to estimate the last-saved quality level, it can be compared against any quality values stored in the metadata.

If the derived quality matches the metadata, then it could be that the metadata reflects the last save. Sign up using Facebook. Sign up using Email and Password.

Post as a guest Name. Email Required, but never shown. The Overflow Blog. Podcast Making Agile work for data science. Stack Gives Back Featured on Meta. New post summary designs on greatest hits now, everywhere else eventually.

Related 2. Hot Network Questions.