Camera phones; what's the quality expectation?

Continued from 'Cell Phones: Excessive Complexity?'….

I thought a lot about the controversy surrounding the future of cell phones when I met with Nethra Imaging in late April, as the company prepared for this past Monday's unveiling of its NI-20×0 image processor family. Targeting up-to-3 Mpixel camera-inclusive cell phones, the base product in the ARM7TDMI-based family is the NI-2050; the NI-2060 variant comes in CSP and costs $8 (10,000 units) while the NI-2080 adds 8 MBytes of SDRAM in a multi-die CSP and costs $12. Sample devices and evaluation kits are now available, and the company hopes to be in production by the end of the third quarter.

Nethra's first-generation architecture supports the following high-level features:

• Color processing
• Detail processing
• Enhanced low light processing (adaptive color or digital dodge and burn)
• Red eye detection in hardware, correction in software
• Electronic image stabilization

Company representatives claim that Nethra's chips deliver "enhanced image quality comparable to high-end digital cameras". I don't doubt that they'll achieve at least most what they assert, among other reasons due to the impressive pedigree of the company's management team, who have previous stints at companies such as Atsana Semiconductor, Chips and Technologies, Cirrus Logic, HP Labs, LSI Logic, MediaQ, NeoMagic, NUCORE Technology, Philips, Pixim, S3 and Toshiba. What I doubt, however, is that an appreciable number of folks are going to be willing to pay for "enhanced image quality comparable to high-end digital cameras".

The feature on Nethra's list that I most resonate with is 'enhanced low light processing'. It reflects the fact that camera phones often get used in bars, restaurants and other dimly lit settings, a phenomenon that I discussed in detail in my recent image sensor cover story. Users expect that their camera phones will deliver an after-the-fact recognizable representation, on the phone's LCD, of the captured scene and the people in it. But an image that they'll make a 5"x7" or larger print of, or post to their website blog in high resolution? If that's their expectation, I predict they'll reach for a camera instead.

I'm not the only one who's skeptical; in the May 2 issue (volume 5, issue 15) of iSuppli's Market Watch newsletter, director and analyst Scott Smyser stated that "baseband processors are evolving from being simply communications processors to becoming more full-featured processors supporting additional features. Standalone application/multimedia processors that support multimedia functions and high-level operating systems previously have been designed into mobile phones in order to handle many of these additional features. However, the baseband processor increasingly is integrating more of the media processing, eliminating the need for a separate application/media processor".

I forwarded the iSuppli report to Nethra and asked for their feedback. Here's what Murty Bhavana, VP of Marketing, had to say; "I’m sure you realize from our discussion that we don’t agree with this perspective. In fact, we are seeing just the opposite. The top tier handset OEMs who we’re working with are all asking for standalone image processors. Current apps CPU vendors have not delivered the image processing capabilities (performance, quality, features and ease of integration) to the handset OEMs. Nethra has RFQs from OEMs who are in their decision making process. These are clearly showing the OEMs open mindset for camera image processors as a standalone chipset. Nethra is currently engaged with all the top Korean OEMs and has begun promotion activities in Japan and USA. Integrating the image processor in a baseband is a good solution when the “camera feature” is a check-off value for the consumer. However, integrating the ISP (Image Signal Processor) in the baseband has other challenges. Baseband, Compression, Display, and ISP technologies have different tracks for evolution. OEMs have faced the challenge of time-to-market and software development and integration for such a full featured product. Nethra believes this architecture will exist in the market, but will always be adopted in the value segment.

"Currently there are at least 10 discrete image processing solutions on the market addressing this segment. Nethra holds an edge in quality of image processing and overall system level solution for the market. TI has an interesting take in this space. They call their solutions – image accelerators and not image processors. Even their highest end DSP is not capable of processing our algorithms in real time, a key requirement for the OEMs. For example, one study on the Web has determined that DSPs are capable of around 10 MPixels/sec type of throughput. This is capable of doing around 3 fps for a 3 MP image sensor. However, OEMs are looking for nearly 15 fps (for current models) type of performance by requesting chipset vendors provide a throughput of nearly 60 MPs/sec (some guard band factored into the number of 54 MPs/sec). Nethra guarantees this performance via their solution and hence the adoption. Yes, at the end of the day the DSPs provide the ultimate in flexibility of imaging algorithms. But the camera module OEMs and ODMs (even in the DSC space) do not have imaging background to implement imaging algorithms."

Let's see what another analyst firm, In-Stat, says about feature-rich phones in its report "Warning: Early-Adopters Have Lukewarm Response to Multimedia Handsets" (#IN0502106WH), published in early May. "Fewer than 9% of respondents to an In-Stat early-adopter consumer survey were very or extremely interested in buying a cell phone capable of playing MP3 or other music files, and less than 11% were very or extremely interested in broadcast TV functionality, the high-tech research firm says. "Cell phone manufacturers, carriers and content providers face serious challenges in convincing these end-users of the benefits of music- or video-centric phones," says Neil Strother, In-Stat analyst."

What do you think about my opinion that cell phones are getting too feature-rich and hard-to-use? Am I a Luddite? Or am I right? And if I am right, what's it mean for the future of the cell phone industry and its symbiotic co-dependents?