Skip to content
Please check an answer for every question.
Cookie settings

Total may deposit the following categories of cookies: Cookies for statistics, targeted advertising and social networks. You have the possibility to disable these cookies, these settings will only be valid on the browser you are currently using.

Enabling this cookie category allows us to establish statistics of traffic on the site. Disabling them prevents us from monitoring and improving the quality of our services.
Our website may contain sharing buttons to social networks that allow you to share our content on these social networks. When you use these sharing buttons, a link is made to the servers of these social networks and a third-party cookie is installed after obtaining your consent.
Enabling this cookie category would allow our partners to display more relevant ads based on your browsing and customer profile. This choice has no impact on the volume of advertising.

Using wavelet packet compression
for seismic data

A project presented by Long Qu, Raphaël Lencrerot, Marianne Cuif-Sjöstrand and Bruno Conche

Long Qu

Best Innovators

Our Pangea supercomputer gives us major high performance computing (HPC) resources that are essential to geoscience studies. However, seismic data processing performance does not rely on computing power alone. The speed with which the processors and file systems read and write input and output data is also important. For this reason, data compression is critical for increasing Pangea’s performance. This understanding led us to adapt wavelet packet compression (WPC) to seismic data, in a patented innovation that optimizes processing speed.

WPC, an Inexpensive Compression Method

The best way to deal with 2D and 3D oscillating data such as that obtained from seismic acquisition is to convert it into 2D wavelets. We therefore designed and deployed a method based on lossless wavelet packet compression (WPC). The flexible parameters in quantizing wavelet coefficients allowed us to obtain a small error metric and limited data loss. This format saves ten times more storage space.

Widely applicable for improving compression speed, even with large volumes of data, this method can be likened to such well-known formats as MP3 for music and JPEG for pictures. This type of compression with very little loss comprises three key stages: conversion, quantization and coding. Decompression follows the opposite path.

Using Blocks to Optimize WPC to Boost Performance

Although conventional WPC suited our use, compression speed still wasn’t fast enough to accelerate the input/output data flow. We therefore looked for a way to adapt wavelet conversion to the architecture of modern processors. In WPC, this computing takes up 80% of compression time. Multi-core processors have several levels of cache memory, with an extremely fast access time but a limited storage capacity of just a few megabytes. Because the data needed for conversion exceeds the cache memory’s size, it has to be reloaded, which means a great deal of back and forth between the processor and the main memory. Instead of converting all the data into wavelets, we decided to innovate by dividing it into small blocks so that the data in each block could fit into the processor’s cache memory.

This patented design avoids reloading data from the main memory, optimizes the reuse of data in the cache memory and considerably reduces computing time without impacting the compression ratio. Combined with other improvements, our solution makes it possible to achieve compression speeds of around 300 megabytes per second and decompression speeds of 400 megabytes per second for wavelet conversion at three levels of cache memory on a single Pangea processor core, which is nearly 25 times higher than conventional WPC.

Many Advantages

Enhanced WPC is a technological breakthrough that considerably limits the use of file systems, accelerates processing speed, optimizes HPC infrastructure and reduces the cost of seismic studies. This represents a significant advantage for our numerous seismic study projects.

  • WPC makes it possible to use beam migration without impacting quality. The geoscience models and structural interpretation of the Louro deep offshore field in Angola’s Block 32 were done this way, with WPC reducing the volume of input data by a factor of ten. The beam migration code can therefore benefit from WPC and help reduce the total volume of input data from 960 terabytes to 25 terabytes, at least three times faster than standard code.
Difference on the final stack from BEAM migration with normal data and compressed data - WPC Compression of seismic data - Best Innovators 2019 - Exploration-Production - Total


  • WPC accelerates 3D full waveform inversion (FWI) and reverse time migration (RTM) tools. At present, 90% of Pangea’s time is devoted to seismic imaging, and 60% of that is focused on RTM and FWI. The use of WPC for these two leading-edge technologies can also reduce the computing time and duration of seismic processing. This could mean savings of up to 35% with the Pangea supercomputer based on graphics processing unit (GPU) technology.
Best Innovators 2019 - WPC Compression of seismic data - Exploration & Production - Total

Best Innovators

The 10 Innovations Rewarded in 2019


Pangea III : Computing Power Breaking New Ground

Research & Development

Delve into 60 Years of Invention in Oil & Gas