Jeffrey Powers

Jeffrey Powers

39p

56 comments posted · 5 followers · following 0

448 weeks ago @ Torbit blog - WebP statistics · 2 replies · +1 points

Hey Josh - An interesting chart, but that blog's analysis is actually wrong. If you study the graphs, you can see that it actually shows JXR performing worse then JPEG (!?). Obviously this isn't true. The problem is that WebP uses 4:2:0 color encoding (so, only 25% of the color is retained), whereas this guy's invocation of JPEG XR retained full color planes (4:4:4) (or, retaining 100% of the color channels). You'd have to re-do the comparison using 4:2:0 to compare apples to apples.

See the later comments on the post for more on the formats used.

But as I said, in my careful analysis, WebP is indeed a bit superior in total compression, but XR would be neck-in-neck with the WebP line on that chart and not below JPEG. But even being slightly superior to XR, WebP still isn't the best still image compression out there!

448 weeks ago @ Torbit blog - WebP statistics · 5 replies · +1 points

Hey guys -- I can add a little to the discussion from personal experience.

Quality:
In my tests, in terms of raw compression abilities, WebP edges out JPEG XR, and both beat JPEG.

Development activity:
There's a guy who is constantly working on the WebP codec (optimizing it for SSE, etc), which is a nice tight open source package.

In contrast, there doesn't appear to be any active open source development of JPEG XR these days, which is a shame. However, JPEG XR has a nice free C standard reference implementation available with a perpetual license.

Breadth:
It's completely true that the WebP format is really limited at this point. It does precisely one thing very well -- encodes photographs for human viewers -- similar to the way people usually use JPG (4:2:0 mode). JPEG XR, however, is incredible breadth -- it could legitimately replace all of JPG, PNG, and RAW formats. WebP has experimental alpha support now, but needs to catch up, badly.

Speed:
JPEG XR is _much_ faster than WebP in encoding. I'm not sure how they differ on decoding. But sometimes you don't care how long the encode takes -- you just want the best compression your network bits can buy. Now that WebP's on the block, it would be nice to see a JPEG XR iteration that had an expensive encoder option which could achieve better bitrates.

Conclusion:
There's no silver bullet. But supporting both of these formats depending on what browser's on the other end would be a great way to speed up the internet :)

483 weeks ago @ Occipital - Announcing ClearCam 1.0 · 0 replies · +1 points

Wow, really? Hadn't read that. Does sound like a job for ClearCam. As for GPU acceleration - very good idea.

493 weeks ago @ Feld Thoughts - eBay Acquires RedLaser · 0 replies · +1 points

Thanks Scott! It was pretty painful at the time, but so far we've been lucky in turning failures around. The new additions to the team are definitely world-class, not to mention two of them happen to be sourced from around the world, too :)

493 weeks ago @ Feld Thoughts - eBay Acquires RedLaser · 3 replies · +1 points

Thanks for the great post on us and our bootstrap story. Very accurate, but I'd add that roughly half of our trajectory so far has been from decisions with ample alternative routes. By and large, those decisions were the product and technology decisions. But as far as running the business itself goes, we've been flying by the seat of our pants, carving out the only path that would keep the company alive. We genuinely did want to raise some money post-Techstars, but the options for doing so were next to nil. Failure was not an option, so we had no choice but to engineer our way to profitability despite the sexier things we really wanted to engineer. Now we're finally back to where we were in the fall of 2008 with an obnoxiously difficult project ahead of us, but this time a little less naive and a little more proven.

501 weeks ago @ Occipital - Create RedLaser Custom... · 0 replies · +1 points

Unfortunately not, we're waiting until we can run native code on the blackberry because Java is too slow to handle RedLaser's algorithm.

516 weeks ago @ Colorado Startups - Occipital solves anoth... · 1 reply · +1 points

Indeed it's a very hard problem! You should consider using the RedLaser algorithm via the SDK. There's a license cost, but it will save you tons of frustration trying to solve it yourself. http://redlaser.com/SDK.aspx

520 weeks ago @ Feld Thoughts - RedLaser #2 Paid App i... · 2 replies · +1 points

You should check out RedLaser custom apps - http://redlaser.com/apps . You could quite easily make a custom app that sends barcodes to a mobile Safari costpad.com page. It's all URL-based, so users just need RedLaser installed, nothing else.

520 weeks ago @ Occipital - Create RedLaser Custom... · 0 replies · +2 points

There's not currently a way, but we'll take this as a feature request for our next update!

521 weeks ago @ Occipital - Announcing RedLaser 2.5 · 0 replies · +1 points

Mugged - If you're seeing US stores, this seems like a likely configuration problem. Are you seeing prices in USD? If so, go to your home screen, then Settings, then scroll down to RedLaser and switch the Currency to GBP.

If you're seeing USD prices, but US stores, then go to Settings again, select General, then International, and check your Region Format and make sure it's set to United Kingdom.

Please email info@occipital.com so we can assist you directly.